Proceedings of the Sixth International AAAI Conference on Weblogs and Social Media
Docs.com: Social File Sharing in Facebook
Shelly D. Farnham, Andrzej Turski, and Sameer Halai
FUSE Labs, Microsoft Research
{Shellyfa, andrzejt, sameerh}@microsoft.com
Abstract
Social file sharing systems that enable public or broadcast
file sharing have been examined in the enterprise space.
For example, a study of the file sharing system Cattail deployed internally at IBM found that adding sociality to file
sharing improved document reach and discoverability
(Shami, Muller and Millen, 2011). Users employed the social aspects of the system to upload and publicize their files,
annotate and track other people’s files, and discover and
share new files (Muller, Millen, and Feinberg, 2010).
However, the use of personal SNSs for file sharing has
yet to be studied. We might expect that users would not
care to share Microsoft Office documents in personal SNSs,
because they tend to be used for more productivity-oriented
activities. Further, users may prefer to not share to their
whole network, but rather only to a few friends, for which
other tools such as email would suffice. To examine these
issues we analyze the content of documents users choose to
share in Docs.com, how they use SNS settings to define
sharing and editing levels, and what impact such social file
sharing has on engagement and document reach.
Social networking sites provide a unique opportunity for users to engage in broadcast online file sharing and collaboration. In this study we examine the usage of Docs.com,
which integrates Microsoft Office web documents with Facebook. In reviewing the content of Docs.com documents,
we found they were equally created and shared for work,
personal productivity, and play. Users actively employed relationship levels in their Facebook networks to define access
to their documents. However, we also found the most
viewed documents were those shared to the entire public,
with Facebook notifications driving most of the traffic to
these public documents. Finally, through a factor analysis of
user activity four primary types of usage emerged – sharing,
creating, collaborating, and consuming. Those who created
and collaborated around documents had higher levels of engagement than those who just shared or consumed documents, interacting with more documents and returning on
more unique days.
Introduction
Social networking sites (SNSs) are increasingly prominent
tools for maintaining both close and distant social relationships. A recent Pew Internet and American Life report
found that 59% of Internet users have at least one SNS, out
of which 92% have a Facebook account (Hampton et al.,
2011). Facebook users have friended 48% of their real
world extended networks, including friends, coworkers,
family, neighbors, and old classmates.
SNS features such as broadcast updates, news feeds, and
activity notifications enable a more lightweight, passive
form of interaction than found in other forms of computermediated communication such as email (Farnham, 2008;
Hampton et al, 2008; Wohn et al, 2011). In the following
note we explore the extent to which users will leverage this
more passive context for social file sharing and online collaboration. In particular, we examine a system called
Docs.com, which provides tools for sharing and collaborating with Microsoft Office documents such as Word, Excel,
and Powerpoint in the context of Facebook.
Docs.com
Docs.com is a web application that allows users to create or
upload Microsoft Office documents online and then share
them with friends using Facebook. Docs.com was released
in June of 2010, and through steady linear growth over the
course of 15 months has accrued 945,000 authenticated users and 298,000 documents. The primary document types it
supports are Microsoft Word, Excel, Powerpoint, and
PDFs. Users sign in to Docs.com using Facebook’s registration and login system (Facebook Registration API,
2011). Once signed in, users may see, read and comment
on either public documents in the document gallery or any
documents their Facebook friends have shared. More importantly, users may create or upload their own online documents. They may edit these documents directly online, or
open them in a local client to edit. When viewed online in
Docs.com, the document is framed in the context of social
sharing features. See Figure 1. The user may set sharing
and editing levels within the Facebook network, including
Copyright © 2012, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
451
a) “Individual friends”, where the user specifies which Facebook friends have access, b) “All my friends”, which includes all of the user’s Facebook friends, c) “Group”, or
any member of a Facebook group to which the user belongs, d) “Everyone”, the entire Internet, and e) “Only me”,
viewable only by the user. At the time of sharing a document, a notification with a link to the document is sent to
the news feeds of appropriate friends or groups. The user
may also add tags to the document, and viewers or editors
may add comments on the document’s wall.
documents were Word, 18% PDFs, 15% Powerpoint, and
15% Excel. Amongst the top 2000 most viewed documents, PDFs were more popular at 35%, with Word at 35%,
Powerpoint at 15%, and Excel at 15%.
User Action
N
Percent
Join the system
75803
9.6%
Create document
41856
5.3%
Upload document
35339
4.5%
Start editing document
79140
10.1%
Save while editing document
56506
7.2%
Finish editing document
60571
7.7%
Share document
14758
1.9%
View document
322656
41.0%
86787
11.0%
Download document
Delete document
Total
12784
1.6%
786200
100.0%
Table 1. User actions collected during data collection period.
Types of Content Shared
In order to assess the types of documents users were sharing
in the context of Facebook, we performed a content analysis
of 400 public documents. Note 32% of all the documents
used were public. We excluded documents that had extremely high numbers of views and thus were not representative of the population (with z-scores over 5.0). We included for analysis 200 of the most viewed documents, and
200 documents selected using a random number generator.
All documents were reviewed twice, first to code for specific, descriptive types of use (e.g., recipe, or political essay),
and then to categorize them into the more general types of
content sharing that emerged. As can be seen from Table
2, the most prominent category of content was fun and entertainment (44%), with work-related documents a close second (38.7%). A third category was personal productivity
(11.3), which contained content relevant to accomplishing
non-work life tasks such as organizing finances or finding a
health clinic. The top viewed documents were less fun and
had more advertisements than those of our random sample.
Figure 1. A document in Docs.com may be viewed and edited
online, and then shared with Facebook friends.
Method
To analyze Docs.com usage, instrumentation data was collected from May 28 to August 7, 2011. The system logged
discrete user actions around documents as listed in Table 1.
For each action, we collected who performed the action,
what document the action was performed upon, the owner
of the document, and attributes specific to each action. For
example, for sharing a document, the system logged whether it was shared only with friends or publicly. When users
viewed a document, the system only logged when the user
first opened the document, not any subsequent clicking.
Use of Sharing Levels
We examined the sharing levels of the 74,671 documents
that were initially created or uploaded during our data collection period. About 45% of these documents were shared
for viewing, and 18% were shared for editing, meaning
55% remained private. The most popular form of sharing is
to everyone, followed by sharing to groups and sharing to
all friends. This suggests that if users share their documents in Facebook, they are casting a fairly broad net. See
Results
Over the 72-day time span of our data collection period
there were 86,897 authenticated users interacting with
104,199 unique documents. 72% of the documents were
new to Docs.com during this time span, with 54% being
newly created online, and 46% being uploaded. 52% of all
452
Table 3. If documents are shared to all friends, they are
shared to an average of 236 friends in the user’s network.
If they are shared to specific people, they are shared to an
average of 1.6 people, out of a maximum of five.
Content
Categories
Fun and entertainment
Work
Personal
productivity
Civic participation
Content
Examples
recipes, sports statistics, writing,
hobby how-tos
presentations, resumes, homework,
schedules
computer safety
tips, health information, forms
call to actions, political opinion articles
Percent
in Random
Sample
Percent
in Top
200
Viewed
44.0%
21.4%
38.7%
41.7%
11.3%
16.5%
4.2%
4.9%
Advertisement
hotel flyers, product information
1.8%
12.6%
Religious
written essays
0.0%
2.9%
can be seen from Figure 3, the majority of these document
views are generated from Facebook. Thus even the public
documents are leveraging the Facebook network to drive
traffic through notifications in users’ news feeds.
Figure 3. More document views are generated if the document is
shared more broadly in the network.
Table 2. Types of content in public documents.
Sharing Level
To View
To Edit
Everyone
12.7%
Group
14.0%
6.2%
Only friends
12.4%
1.2%
Specific people
4.2%
2.2%
Page admin
1.1%
7.7%
55.6%
82.7%
Only Me
Figure 4. Most document views for public documents are generated from links within Facebook notifications.
Types of Usage
To enable comparisons with Muller, Millen and Feinberg
(2010) who examined social file sharing in the enterprise,
we performed the same factor analysis method to reduce
our various activity measures to the most important types of
use, first aggregating the data at the level of the user. As
can be seen from Table 4, four factors emerged. Prominent
types of usage tended to be either a) sharing, uploading
documents to share them, b) creating, creating and editing
their own documents online, c) collaborating, viewing and
editing others’ documents, or d) consuming, viewing and
downloading public documents. Distinct from the findings
of Muller, Millen, and Feinberg (2010), creating and collaborating were important components of Docs.com usage
patterns.
Table 3. Percentage of sharing levels across documents for both
sharing for others to view, and sharing for others to edit.
Sharing Level and Document Reach
Document reach was measured as number views by the self,
friends, and unauthenticated public persons. Sharing levels
had a meaningful impact on document reach, as can be seen
from Figure 3. Documents shared with the user’s entire
network (only friends) or a group had higher levels of views
from friends. Documents generated much higher page
views on average when shared to the entire Internet (Everyone). In order to examine whether the page views for these
public documents came from Facebook notifications, we
examined the referring site for each document view. As
Individual scores were then saved for each factor to assess
which types of usage had the most impact on two measures
453
of engagement: the number of unique days the user was in
the system, and how many documents they interacted with.
As can be seen from Table 5, users who were more engaged
with sharing content interacted with more documents,
whereas those who created content and collaborated returned for more days. In other words, the use of the more
collaborative features corresponds with more engagement.
gration with Facebook to engage in broadcast, social file
sharing and online collaboration. We found that Docs.com
is well-used and continues to grow, that users actively used
the network-based access levels to passively share their
documents to all their friends and groups within Facebook,
and that notifications within Facebook drove most of the
traffic to users’ documents, especially for documents shared
to the public. We found four primary types of usage – sharing, creating, collaborating, and consuming – and that those
who created and collaborated around documents had higher
levels of engagement.
Factor
Eigenvalue
2.4
2.3
1.3
1.1
23.0%
22.9%
12.8%
11.1%
Sharing
Creating
Collaborating
Consuming
.032
.783
.046
.037
.940
.121
.018
.015
.035
.922
.044
-.016
-.006
-.085
.887
-.130
.721
.023
.011
-.016
.942
.153
.021
.016
.061
.317
.694
.275
.240
.826
.054
.000
-.024
-.003
.039
.755
.024
.012
-.001
.668
% Variance
Measure
Created doc
Uploaded
doc
Edited own
doc
Edited others' doc
Shared doc
to edit
Shared doc
to view
Viewed
friends doc
Viewed own
doc
Viewed public doc
Downloaded
doc
Although we might expect Microsoft Office documents
to be more productivity-oriented, we did find that a substantial percentage of Docs.com documents had fun and entertaining content, as might be expected from sharing within a
friend-oriented network such as Facebook. Nonetheless,
we also found many documents were focused on work and
personal productivity. All documents, whether for fun or
work, tended to be focused on sharing information. Unfortunately for this study we were only able to examine public
documents, and in the future expect to pursue a comparison
with private documents through questionnaires or interviews.
Our results show that users who effectively use social
media sharing tools can drive increased engagement with
their online content, and that SNSs provide a unique context
for a form of lightweight micropublishing – to the user’s
extended network.
Table 4. Factor loadings for types of usage. Sharing, creating,
collaborating, and consuming emerged as four types using principal component analysis with Varimax rotation. We considered
any cell with a value > .40, as shaded.
Unique Days
Active
Number
of docs
r
r
Sharing
.20
.64
Creating
.50
.38
Collaborating
.44
.29
Consuming
.13
.23
Type of Usage
References
Facebook
registration
API.
http://developers.facebook.com/docs/plugins/
Farnham, S. D. 2008. The Facebook Application Ecosystem:
Why Some Thrive – and Most Don’t. An O’Reilly Radar Report,
March 2008.
Farnham, S. D. and Churchill, E. F. 2011. Faceted Identity, Faceted Lives: Social and Technical Issues in Being Yourself Online.
In Proc. CSCW 2011, ACM Press, 359-368.
Hampton, K. N., Goulet, L. S., Rainie, L., Purcell, K. 2011. Social Networking Sites and Our Lives. A report of the Pew Internet
& American Life Project.
Muller, M., Millen, D. R., and Feinberg, J. 2010. Patterns of Usage in an Enterprise File-sharing Service: Publicizing, Discovering, and Telling the News. In Proc. CHI 2010, ACM Press, 763766.
Shami, N. S., Muller, M., and Millen, D. 2011. Browse and Discover: Social File Sharing in the Enterprise. In Proc. CSCW
2011, 295-304.
Wohn, D., Y., Lampe, C., Vitak, J., Ellison, N. 2011. Coordinating the Ordinary: Social Information Uses of Facebook by Adults.
In Proc. iConference, ACM Press, 340-34.
Table 5. Correlations between primary types of usage and
measures of engagement. Due to the large N, all r values are significant at p < .001, and should be interpreted as effect sizes
where .2 = weak, .5 = moderate, and .7 = strong.
Conclusion and Discussion
In this analysis we examined the usage patterns of
Docs.com to assess whether people will leverage its inte-
454